REPORT

Analysis of groundwater resources in densely populated urbanwatersheds with a complex tectonic setting: Shenzhen, southern China

Michele Lancia1,2 & Chunmiao Zheng1,2& Shuping Yi1,2 & David N. Lerner1,2,3 & Charles Andrews1,2,4

Received: 23 November 2017 /Accepted: 30 August 2018 /Published online: 24 September 2018# The Author(s) 2018

AbstractShenzhen is the major financial and high-tech center in southern China. The megacity has grown rapidly in the last 40 years withthe population increasing from about 30,000 in 1979 to 20 million in 2016. The study area (2,015 km2) is about 42% urban and58% undeveloped land. The rapid development of the megacity has resulted in severe degradation of the groundwater andsurface-water resources and has created a nearly insatiable demand for water, with an average consumption of 2000 × 106 m3/year. Groundwater is an important component of the baseflow of themany streams in the area and is used for potable water supplyand irrigation in some of the rural parts of the municipality. This study develops a conceptual model and quantitative frameworkfor assessing the groundwater resources of Shenzhen. The groundwater system consists of shallow aquifers of alluvium andweathered bedrock overlying low permeability igneous and sedimentary rocks. The complex geologic setting was conceptualizedas a block structure with blocks bounded by high-angle faults. The water budget in Shenzhen was quantified. The estimatedaverage groundwater discharge is about 12% of annual precipitation. The study provides a starting point to investigate how amegacity such as Shenzhen shouldmanage and protect its groundwater as a strategic resource and environmental asset. It is also abasic management tool for analyzing and contributing to urban drainage concepts such as the Bsponge city .̂

Keywords Conceptual models . Numerical modeling . Groundwater recharge/water budget . Urban groundwater . China

Introduction

Shenzhen, in Guangdong province, is an important megacity insouthern China which has experienced a huge and rapid devel-opment. Its area extends along the southern Chinese Sea coast-line between the Zhujiang River (Pearl River) Estuary and DayaBay and shares its southern boundary with Hong Kong (Fig. 1).

The rapid development has caused a large reduction in agricul-tural lands and green areas (Li et al. 2010; Hao et al. 2011).Urbanization increased from 9 to 40% of the city area between1970 and 2008 (Chen et al. 2014; Hong and Guo 2017). Urbanexpansion continues and has reached 42% of the study area, asshown by analysis of satellite images (Fig. 1). Construction ofreservoirs for drinking and industrial water uses (Wanget al. 2004) has greatly altered the morphology of the rivervalleys (Qin et al. 2013). The water supply (2000 × 106 m3/year) is mostly imported (~75%), as the remaining contribution(only ~25%) comes from reservoirs and water-wells distribut-ed inside the study area (WRBSM 2016). The groundwaterresources of Shenzhen are underutilized and the amount ofpumped water is only 5.9 × 106 m3/year (WRBSM 2016).Extensive pollution affects surface waters and the aquifers,with high costs for wastewater treatment systems (Li et al.2016). Notwithstanding the link between economic develop-ment and the environment, the hydrogeological setting ofShenzhen has not been thoroughly investigated.

The earliest available compilation of hydrogeological in-formation was GBGP (Geological Bureau of GuangdongProvince, China, Unpublished report on regional

* Chunmiao Zhengzhengcm@sustc.edu.cn

1 Guangdong Provincial Key Laboratory of Soil and GroundwaterPollution Control, School of Environmental Science andEngineering, Southern University of Science and Technology,Shenzhen, 518055, China

2 State Environmental ProtectionKey Laboratory of Integrated SurfaceWater-Groundwater Pollution Control, School of EnvironmentalScience and Engineering, Southern University of Science andTechnology, Shenzhen 518055, China

3 Department of Civil and Structural Engineering, University ofSheffield, Sheffield, UK

4 S.S. Papadopulos and Associates, Inc., Bethesda, MD, USA

Hydrogeology Journal (2019) 27:183–194https://doi.org/10.1007/s10040-018-1867-2

hydrogeological survey, scale 1:200.000, 1980) which com-piled many borehole logs from throughout Shenzhen alongwith recovery tests and hydrogeochemical data. A decadelater, Yang et al. (1991) described the geology of theShenzhen area, attributing the current structural setting tostrike-slip tectonics with a main element oriented NE–SW.Campbell and Sewell (1997, 1998) reconstructed the differentplutonic stages and their relationship with the strike-slip tec-tonics. Xu et al. (2015) investigated the structural setting froma geophysical point of view. The most comprehensivehydrogeological description of Shenzhen is Yang et al. (2007).

Groundwater is strongly influenced by the relationship be-tween the aquifers and anthropic infrastructure. In urban areas,groundwater sources and pathways are numerous and com-plex. Buildings, roads and other impervious surfaces impedethe natural aquifer recharge, increasing surface runoff, whileon the other hand, leakage from water mains and sewers pro-vide unconventional groundwater recharge (Lerner 2002).Many studies discuss the impacts of urbanization on thehydrogeological setting (Held et al. 2006; Passerello et al.2012; Kruse et al. 2013; Di Salvo et al. 2014). Due to theeconomic importance of the study area, groundwater is ofgreat interest to local water and environmental managementagencies. The key science question is how a megacity such asShenzhen should manage and protect its groundwater

resources as a strategic resource and environmental asset.Because of the lack of literature and its geological and urbancomplexity, this study aims to be a first step to investigate thisimportant key scientific question through:

& Constructing the three-dimensional (3D) distribution ofthe aquifers and defining a conceptual circulation model

& Evaluating the recharge and permeability of thehydrogeological units

& Identifying and quantifying the groundwater resources ofShenzhen

Hydrogeological setting and conceptualmodel

Shenzhen is located on the continental margin of theCathysian block (Zhang et al. 2013) and experienced the cre-ation of a basin and range structure, affected by the complexgeodynamic setting of the Guangdong region and, more gen-erally, of Southeast Asia (Li and Li 2007). In Shenzhen, thestratigraphic basement consists of Lower Paleozoic or olderclastic rocks with low-grade metamorphism (Fig. 2).Overlying the basement, the sedimentary succession consists

Fig. 1 Satellite image of Shenzhen and Zhujiang River estuary (fromCNES-2017). The urban areas were defined bymanual analysis of satellite imageryat 1:5.000 scale

184 Hydrogeol J (2019) 27:183–194

of Upper Paleozoic and Mesozoic clastic sequences character-ized by a continental to shallow marine-water environment.The stratigraphic sequence varies from siltstone to sandstonewith layers of gravel and breccia (Fig. 2). Carbonate sedimen-tation in the Lower Carboniferous is represented by a se-quence of dolostone and limestone, preserved in thePingshan and Longgang area (Fig. 2b). The thickness of thesuccession is large and difficult to characterize in its entirety,due to the heterogeneous sedimentation with several hiatusand transgressive events. Yang et al. (2007) estimate a maxi-mum thickness around 15 km for the whole succession. LateJurassic–Early Cretaceous granitic intrusions related to thesubduction of the paleo-Pacific slab (Fig. 2a,b) are widespreadand interrupt the sedimentary succession (Li and Li 2007; Xiaet al. 2011). This regional magmatic event is known as theYanashanian event (Zhang et al. 2013). As a result of theintrusive activity, the sedimentary succession is intersectedby contact metamorphism, related to two phases of intrusionthat outcrop in the Bao’An and Nanshan areas and along theMirs Bay, Dapeng peninsula (Fig. 2b). An acid magmatismresulted in a volcanic edifice of rhyolitic lava flows and tuffslocated at Wutong Mt. (Fig. 2b). Cretaceous tuffs associatedwith loose ash are scattered along the Dapeng peninsula aswell as in the Pingshan and Longgang basins (Fig. 2).Younger reworked tuff in a fluvial-lacustrine environment alsooutcrops in the Pingshan basin (Fig. 2b). Yang et al. (2007)date these sequences from the Cretaceous to the Eocene.Quaternary deposits are thickest in the main and secondaryriver valleys and along the coast. The thickness of these de-posits varies as a function of the inherited morphology; be-tween 30 and 60 m in the coastal sector and main river valleys,and 15–40m in the secondary river valleys (Lu and Sun 1991).

The distribution of the Paleozoic and Mesozoic rocks iscontrolled by the different tectonic stages that also influencethe trend of the river valleys. The Lower Paleozoic tectonics ischaracterized by contractional faults and thrust planes, NE–SWoriented and NW verging, created during the phases of theCaledonian orogenesis (Pang et al. 2016). During the earlyMesozoic, the orogen experienced a distension associatedwith high rate of subsidence, due to roll back of the paleo-Pacific plate. The study area was intersected by strike-slip tec-tonics (Allen et al. 1997) with the Meixian-Shenzhen fault asimportant tectonic line (Sun et al. 2007). The latter divides theBao’An area, characterized by an Upper Paleozoic hiatus, fromthe Paleozoic sequence of the Longgang-Pingshan area. Thesefaults were active up to the Quaternary and in some cases up tothe Middle part of the Upper Pleistocene (Lu and Sun 1991).

For the conceptual model the great thickness of sedimen-tary units, associated with the strike-slip tectonics, is schema-tized as a block structure, whereby every block is delimited byhigh-angle faults or granite intrusions. Cross-sections in Fig.2b are representative of the regional geological setting. Theliterature, especially the borehole stratigraphy from

GBGP(Geological Bureau of Guangdong Province, China,Unpublished report on regional hydrogeological survey, scale1:200.000, 1980) and the hydrogeological map from Yanget al. (2007), supports a division into eight mainhydrogeological units:

& The Quaternary clastic unit (legend 1 in Fig. 2) consists ofa multi-layered sequence with the hydraulic conductivitystrongly related to the predominant size of the clasts.Deposits are heterogeneous; layers consist of coarse tosilty sands, with intercalations of sandy gravels along theriver valleys or silty clays in coastal areas as the result ofmarine transgressions and/or temporary lacustrine events.In addition, layers of detritus and talus, derived from theweathered bedrock, have been observed. Thishydrogeological unit has a thickness varying between 30and 60m along the main valleys and in the coastal sectors.

& The Tertiary reworked tuff unit consists of loose ash andtuffs (legend 2 in Fig. 2) which belong to the upper portionof the volcanic sequence. This hydrogeological unit alsoincludes reworked deposits with intercalations of coarsegravel observed in outcrop in the Pingshan area.

& The Upper Mesozoic lava unit (legend 3 in Fig. 2) consti-tutes Wuton Mt. in the southern part of Shenzhen. Aninterconnected fracture system characterizes the lavaflows. Individual fracture orientation is directly correlatedto the movement of the flow. Intercalations of paleo-soilsor weathered deposits locally reduce the hydraulicconductivity.

& The Jurassic granite unit (legend 4 in Fig. 2) is an aquitard.Facture density is very low and RQD (rock quality desig-nation) index records excellent values (90–100%), beyond30 m from the ground surface. Towards the land surface, ablanket of more fractured and weathered deposits is pres-ent. This band typically extends to 30 m below the landsurface based on the analysis of the stratigraphy of manywater-wells and boreholes. Usually the shallowest 15-m-thick portion of this unit consists of a very permeablecoarse quartz sand. This shallowest part represents themain aquifer, where the groundwater circulation is moredeveloped. In the eastern sector, the granite aquitard(legend 4a in Fig. 2) is more fractured (Yang et al. 2007)but with a thinner weathered band, gravitationally mobi-lized because of the steep morphology. From ahydrogeological point of view, this granite (legend 4a inFig. 2) is grouped with the Upper Mesozoic lava unit.

& The Upper Paleozoic–Mesozoic unit (legend 5 in Fig. 2)consists of a sequence of quartz sandstone, siltstone, sandysiltstone and shale. Deposits are widely intersected bytectonic sediments. Sediments are lithified and the primaryporosity is very low even in the coarser deposits. Fracturesare abundant but the apertures were strongly reduced bydynamic and regional metamorphic processes as proven

Hydrogeol J (2019) 27:183–194 185

186 Hydrogeol J (2019) 27:183–194

by the presence of andalusite, chlorite and other metamor-phic minerals (Yang et al. 2007).

& The Carboniferous carbonate unit (legend 6 in Fig. 2) isstrongly karstified and it is present at the bottom of themain Quaternary depressions in the Pingshan andLonggang areas. Numerous sinkholes occur where thisunit is present. Sporadic outcrops also suggest metamor-phic facies of marble, dolomitic marble and crystallinelimestone.

& The Paleozoic unit (legend 7 in Fig. 2) consists of meta-morphic sandstone, phyllite and slate with serpentine.Porosity is very low due to the huge lithostatic load thatthese rocks experienced. Due to the different tectonic cy-cles, fracture systems are interconnected and more devel-oped than in the Upper Paleozoic–Mesozoic unit. TheBasement unit (legend 8 in Fig. 2) is metamorphosedLower Paleozoic rocks of quarzitic sandstone and muddyshale. Porosity is very low and fracture apertures werestrongly reduced by metamorphic processes. From ahydrogeological point of view, this unit is an aquitardand is grouped with the Upper Paleozoic–Mesozoic unit(legend 5 in Fig. 2).

The stratigraphy is summarized as loose Quaternary clasticaquifers superimposed on old Paleozoic and Mesozoic bed-rock aquitards, covered by a layer of alluvium and colluvium.To illustrate the resulting hydrogeological conceptual model,three sections have been drawn. Figure 3 represents the shal-low groundwater flow paths. Eluvium and colluvium, com-bined with Quaternary deposits, constitute the main aqui-fer, recharged in the rural areas by rainfall. Additional re-charge occurs across the urban areas, in accordance withLerner (Lerner 2002); Fig. 3a,b). Groundwater flows to-ward and discharges into rivers, reservoirs or directly tothe ocean as suggested by Wang et al. (2017). The hydro-geology of the Dapeng peninsula (Fig. 3c) also has a thinaquifer overlying a bedrock aquitard, with recharge fromrainfall and urban sources.

The Lower Carboniferous carbonate unit is an exception tothis conceptual model; it is the most permeable unit of the

study area but it is only found in the bottom of thePingshang and Longgang River valleys, buried by theQuaternary aquifers, as shown in Fig. 3b. As a result of graniteintrusions there is no continuity of the karst unit from onehydraulic basin to the next. Despite the high hydraulic con-ductivity and the ongoing karstification, evidenced by sink-holes (Yang et al. 2007), there is little groundwater movementin this hydrogeological unit except for local flows caused bygroundwater pumping.

In this conceptual model, groundwater and surface water-sheds coincide and thus groundwater circulation can be stud-ied at the catchment scale. Bedrock is essentially an aquitardbut the hydraulic conductivity can be locally altered by thenumerous faults. Across the fault plane, the cataclastic bandsincrease the hydraulic conductivity, functioning as a conduit.On the contrary, the fault zones with a mylonite core havesignificantly lower permeability (Caine et al. 1996). The ef-fects of faults on groundwater flow are small and localized, asobserved by Ran et al. 2013 for normal displacements andCilona et al. 2015 for strike-slip systems. In the MazhouRiver basin (Fig. 2), a single thermal spring illustrates upwardheat migration from the bedrock to the shallow zone along afault, affecting the local groundwater path. The thermal occur-rence is isolated and favored by the cataclastic band; however,there is no impact of the faults on the bedrock circulation at theregional scale.

Methodology

Construction of the numerical model

A steady-state groundwater model of Shenzhen was devel-oped using the finite difference code MODFLOW(Harbaugh 2005) and the graphical user interface GMS. Thesubsurface was discretized with fivemodel layers (Fig. 4a), setas convertible (GMS 2017), with each layer containing110,000 cells, with dimensions of 500 m in the east–westdirection and 100 m in the north–south direction. TheUpstream Weighting (UPW) flow package and the Newtonsolver (NWT) were used (Niswonger et al. 2011). The top ofthe first layer is the land surface, derived from a digital eleva-tion model (DEM) (BIGEMAP database), with a 2.0 m reso-lution, while the bottom of the first layer is specified as 30 mbelow land surface in areas where Quaternary, granite, volca-nic and sedimentary hydrogeological units are present (Fig.4b). The bottom of the second layer was set at 60 m belowland surface. This layer represents the deeper Quaternary ba-sin of the Longgang area, the coastal area of Bao’An, Nanshanand Futian districts as well as the tuff banks in Dapeng. Thebases of the third, fourth and fifth model layers were set atelevations of −150, −300 and − 500 m a.s.l., respectively. Thelatter layers were used to characterize the distribution of the

Hydrogeol J (2019) 27:183–194 187

�Fig. 2 a Geological and hydrogeological setting of the westernGuangdong Province, b the Shenzhen city with sections, and c a reviewof aquifer characteristics. Key to the legend: (1) Quaternary clastic unitfrom coastal plain and alluvial valleys; (2) Tertiary reworked tuff unit; (3)Upper Mesozoic acid lava flow unit; (4) Jurassic granite unit, more frac-tured (a) or more compact (b); (5) Upper Paleozoic-Mesozoic sedimen-tary unit; (6) Carboniferous carbonate unit; (7) Paleozoic sedimentaryunit; (8) Lower Paleozoic (?) basement unit; (9) normal, contractionaland/or strike slip fault; (10) spring, number indicates the elevation ma.s.l. (a), hot spring numbers indicate the elevation and the temperature(b); (11) main river; (12) main reservoir, number indicates the averagewater level; (13) main domestic water-well or piezometer, number indi-cates the elevation m a.s.l.; (4) hydrogeological section; (15) limit of thestudy area of Fig. 2b (represented in Fig. 2a)

karst aquifer at depth. The bottom of the model domain is−500 m a.s.l., a depth at which groundwater movement isnegligible at a regional scale. Geological structures such asfolds and monoclines are represented, due to the primary im-portance of the karst aquifer. Hydraulic conductivity and re-charge rates were specified for each cell based on thehydrogeological unit mapped at the location of the cell.

Constant head boundary conditions (IBOUND = −1,Harbaugh 2005) were specified along the coast with a hydraulic

head equal to zero. Rivers have been represented via the riverboundary package (RIV 1, Harbaugh et al. 2000). River stagehas been derived by DEM. Along the river beds, values of Carc

(conductance for unit length, GMS 2017) vary between 0.02and 0.0001 m2/s, as a function of the width of the rivers andassuming the hydraulic conductivity of the river sediments equalto 0.0001 m/s. The unsaturated zone flow package (UZF) wasused to represent the upper model surface (Niswonger et al.2006). The package can represent thick portions of unsaturated

Fig. 3 Hydrogeological conceptual model of the Shenzhen area,described via three hydrogeological sections through a the ZhuijianRiver Estuary-Guanglang R., b Pingshan-Longgang and c across theDapeng peninsula. Section locations are represented in the mini-map(d). Sections are not to scale to emphasise the hydrogeological circulation

through the shallow eluvium/colluvium soils, the Quaternary clastic se-quence and the tuff layers. Recharge is from rainfall and urban sourcesand groundwater directly discharges towards the reservoirs, rivers or thesea. Granite and the sedimentary successions are aquitards; thus, hydrau-lic catchments and hydrogeological basins coincide

188 Hydrogeol J (2019) 27:183–194

soils, as it is typical for rough and steepmorphologies. Rechargerate is the input variable of this package and depends on thehydrogeological units that outcrop. The Brooks and Corey(1966) model has been assumed constant to describe thegroundwater flow in the unsaturated soils. Epsilon (EPS) hasbeen set equal to 3.5 and the water content of the unsaturatedzone (THTS) to 0.33 (GMS 2017).

The northern boundary of the model consists of many sub-catchments that perfectly follow the watershed lines, as for theGuanlang and Pingshan river valleys. Because the groundwa-ter does not flow beyond the watershed boundary, no-flowboundary conditions (IBOUND = 0, Harbaugh 2005) werespecified (Fig. 4c). Nevertheless, the Longgang River catch-ment extends beyond the boundary of the studied area. Here,the limit of the studied area cuts across a granite mountainsection with narrow gorges. Because of the particularhydrogeological setting across the boundary, continuous aqui-fers are not present. In accordance with the constructedhydrogeological conceptual model, no-flow boundary condi-tions (IBOUND = 0, Harbaugh 2005) have been extendedalong the north-west boundary (Fig. 4c).

However, the model boundary cuts across the LonggangRiver Valley and the groundwater level varies with a

horizontal gradient. The general head boundary (GHB) pack-age (Harbaugh et al. 2000) has been set to better define thegroundwater setting of the area (Fig. 4c). The conductanceterm of the GHB package has been calculated, analyzing theelevation of the river and its distance from the boundary cells.

The boundary between Hong Kong and the study area fol-lows the Shenzhen River and the Shatoujian stream (Fig. 2).The downstream portion of the Shenzhen River has been in-cluded in the River package. The upper stream portion of theShenzhen River and the Shoutoujian Stream are less importantand characterized by a seasonal discharge. This area is char-acterized by volcanic lavas with a shallow groundwater flowdirected towards these two rivers. Hydraulic exchanges be-tween Hong Kong and Shenzhen do not occur and the no-flow boundary condition (IBOUND = 0, Harbaugh 2005) isappropriate.

Model calibration

Homogeneous hydraulic conductivity and recharge were as-sumed for each hydrogeological unit. Analysis of recoverytests of water wells (Geological Bureau of GuangdongProvince, China, Unpublished report on regional

Fig. 4 a Three-dimensional grid of Shenzhen model, displaying the ele-vation (m a.s.l.); b schematic discretization of the grid covering fivelayers; c boundary conditions of the Shenzhen area (map and section).Key to the legend: (1) constant head (IBOUND= −1) for the sea level(h = 0 m a.s.l.); (2) river (RIV package); (3) no flow boundary

(IBOUND = 0); (4) recharge from rainfall and urban areas viaUnsaturated Zone Flow package (UZF); and (5) General HeadBoundary Condition package (GHB), in which the conductance has beencalculated assuming an average hydraulic conductivity, in accordancewith the conceptual model

Hydrogeol J (2019) 27:183–194 189

hydrogeological survey, scale 1:200.000, 1980) showed awide range of hydraulic conductivities. For eachhydrogeological unit, a starting value was set at the averageof the recovery test results. Recharge in undeveloped areaswas based in the first step of calibration on average rechargerates from Yang et al. (2007), with a zero value for urbanrecharge. Recharge and hydraulic conductivity for each blockwere calibrated using a trial-and-error procedure.

The manual calibration investigates how changes in modelparameters vary the fit between simulated values and targets.The calibration targets were 158 water-wells and piezometers.The data were collected via a classic water level meter inMarch 2017, at the end of the dry season, and in July 2017,during the wet season. The two datasets typically differ by 3 min the rural areas and less than 1 m in the flat urban areas.Calibration started by varying the hydraulic conductivity ofthe Quaternary aquifers that outcrop in the coastal plain and inthe alluvial aquifers. Next, the conductivity of weathered soilsand then the bedrock units were calibrated. Initial results weresatisfying from a qualitative point of view but simulated hy-draulic heads differed from the observed by up to 30 m, indi-cating that recharge needed to be included as a model calibra-tion parameter.

For the rural areas, recharge values were constrained to theliterature range of Yang et al. (2007). In urban areas, simulatedhydraulic heads were lower than the observed. Therefore un-conventional water recharge was considered (i.e. leakage fromwater mains, sewer pipes, septic tanks and storm drains(Lerner 1990, 2002). For urban areas, a best fit was foundfor a recharge of 250 mm/year. The final calibrated hydraulicconductivity and recharge datasets are shown in Table 1. Allthe values are in accordance with the conceptual model andthe literature ranges. The calibration plot shows an excellentdistribution of the simulated versus the observed hydraulicheads (Fig. 5a), with a residual value ranging ±6 m (Fig. 5b).

However, calibration of a ground model does not usuallyyield a unique solution. A reduction in hydraulic conductivitywill raise the simulated heads and can be counterbalanced by areduction in recharge. Therefore, a sensitivity analysis wasperformed; the entire hydraulic conductivity and rechargedatasets were varied by ±10%. An increase in the error wasobserved in both cases, without varying the groundwater flowdirection. This informal sensitivity analysis gives some confi-dence that the numerical model is satisfactory.

Results and discussion

Numerical results

The model calibration results are excellent considering themulti-catchment scale and a conceptual model with homoge-neous parameter values for each hydrogeological unit. In the

Zhuijiang River Estuary sector, the flat morphology, the near-ness to the coastline and the high hydraulic conductivity of theQuaternary deposits depresses the hydraulic heads to less than15 m a.s.l. (Fig. 6). The hydraulic gradient steepens to 5% inthe piedmont transition between the coastal plain and the lesspermeable sedimentary and intrusive rocks. Simulated hy-draulic heads match the elevation of some springs in theBao’An area (Fig. 2b). The highest hydraulic heads are ob-served under Yangtai Mt. (maximum head of 150 m a.s.l.) andin the Yangtian Mts. area (maximum head 192 m a.s.l.).

From this area of maximum hydraulic heads, the simulatedgroundwater fluxes are toward the NNW into the MaozhouRiver basin and NNE to the Guanlang River (Fig. 6).Shenzhen Bay receives groundwater from the granite intru-sions of the Yangtai Mt. and Futian area, via fluxes directedfrom North to South. In the Shenzhen River basin (Fig. 6), theflatter morphology and the clastic morphology contribute toflatter gradients, with a variation of 50 m in 15 km (averagehydraulic gradient of 0.3%). In the southern part of the catch-ment, groundwater directly discharges to the Mirs Bay. Thehydraulic head variation is related to the topography, with agradient varying from 3 to 5%. Fluxes are directly perpendic-ular to the coastline and are strongly influenced by the fluvialmorphology. The groundwater of the Dapeng peninsula dis-charges into the Daya Bay in a similar way. The irregularcoastline favors the discharge of the groundwater directly tothe sea, via hydraulic gradients from 0.3 to 1% and hydraulicheads typically between 0 and 30 m a.s.l. Finally, thePingshang and Longgang river basins have similar settingswhere groundwater discharges directly to the river valleys.Hydraulic gradients vary strongly as a function of the predom-inant lithology along the flow path and range from 0.1% in thealluvial aquifer to 4% in the granite. Carboniferous karstifiedlayers, located at the bottom of the Pingshang and Longgangvalleys, are fully saturated, with a groundwater flow that re-flects the shallow groundwater paths.

As expected from the hydrogeological conceptual mod-el, the simulated groundwater basins coincide with the sur-face watersheds. The estimated groundwater balance issummarized in Table 2, where ‘average groundwater dis-charge’ is the annual flux directed into the rivers, reser-voirs or the ocean, according to the black arrows of Fig.6. The model simulations demonstrate that urban waterrecharge is an important part of Shenzhen’s water balance,consistent with Lerner (1990, 2002).

Limits and uncertainties of the numerical analysis

The conceptual model incorporates the most plausible geolog-ical and hydrogeological settings, based on the stratigraphicand the structural analysis from borehole data. All the avail-able information has been used for the model construction andan uncertainty analysis of alternative conceptual models

190 Hydrogeol J (2019) 27:183–194

(Reefsgaard et al. 2006) cannot be performed. The parametervalues of the calibrated model are consistent with the literatureranges indicated by GBGP (Geological Bureau of GuangdongProvince, China, Unpublished report on regionalhydrogeological survey, scale 1:200.000, 1980) and Yanget al. (2007), and are in accordance with the aquifer andaquitard definition of the hydrogeological conceptual model,and have hydrosense (Hunt and Zheng 2012); however, many

combinations of parameter values can give similar results ingroundwater modelling. A sensitivity analysis was performedthat showed that small variations in hydraulic conductivityand recharge (±10%) slightly increase the error in head cali-bration but do not change the groundwater flow paths, where-as, on the contrary, the hydraulic head distribution is not sen-sitive to the conductance specified in the river package. Thestudy has the objective of identifying and quantifying the

Fig. 5 Calibration plots showinga simulated vs. observedhydraulic head and b the residualvs. observed hydraulic head

Table 1 Hydraulic conductivityand recharge datasets used for thecalibration of the numericalmodel

Hydrogeological unit Hydraulic conductivity[m/s]

Recharge rate[mm/year]

Quaternary clastic unit (legend 1 in Fig. 2) 1 × 10−4 300

Tertiary reworked tuff unit (legend 2 in Fig. 2) 1 × 10−3 315

Upper Mesozoic lava (legend 3 in Fig. 2) andfractured Jurassic granite unit (legend 4a in Fig. 2)

4 × 10−7

5 × 10−6 (if weathered)

190

Jurassic granite unit (legend 4b in Fig. 2) 1 × 10−7

5 × 10−5 (if weathered)

150

Upper Paleozoic-Mesozoic sedimentary unit

and Lower Paleozoic (?) basement

(legends 5 and 8 in Fig. 2)

6 × 10−7

5 × 10−6 (if weathered)

70

Carboniferous carbonate unit (legend 6 in Fig. 2) 1 × 10−3 No outcrops

Paleozoic sedimentary unit (legend 7 in Fig. 2) 8 × 10−7

5 × 10−6 (if weathered)

240

Urban Areas – 250

Hydrogeol J (2019) 27:183–194 191

groundwater resources at a regional scale from the currentconstrained dataset. Many limitations are inherent from theregional character of this study and a formal uncertainty anal-ysis of the numerical results (Refsgaard et al. 2007) is notpossible. Parameter values are suitable to describe the regionalflow paths; downscaling to analyze a single catchment,though, would require a new calibration process. Shenzhenis one of the pilot areas for the Chinese BSponge CityInitiative^ (Shenzhen Committee 2016), involving the intro-duction of LID (low impact development) elements to storerainfall in the subsoil and ease its reuse (Chang et al. 2018). Inorder to assess the interaction between urban drainage andgroundwater, the model would require finer discretizationand the inclusion of a sewer network as a boundary condition.

Despite the high number of well-distributed hydraulic headmeasurements, there is a lack of measured hydraulic heads inthe areas with the highest heads near Yangtai Mt. and theYantian Mts. In these areas, the simulated hydraulic headsuggests a thickness of unsaturated granites from 400 to800 m. Supplementary hydraulic heads in these areas

would help the calibration process by providing more dataat locations away from the boundary conditions (i.e. mainrivers and coastline); nevertheless, since the granite is anaquitard, changing the hydraulic head maxima would notinfluence the flux directions.

The diffuse urbanization and the construction of numerousreservoirs make unreliable the calibration of the model viadischarge measures. Indeed, the estimation of the base flowcannot be performed in the areas located downstream the res-ervoirs, as here, the flux is controlled by the water authoritiesand can also vary in a small lapse of time. Obviously, reliablebaseflow data also cannot be inferred in the urban areas. Datacan be collected just in the rural portion, where small sub-catchments close to the watersheds have not been urbanized.Discharge measures would not be sufficient to homogeneous-ly cover the study area and provide a reliable calibration forthe whole model.

Discretization also creates inaccuracies, and more cellswould better define morphology as well as the geometriesof the aquifers. The top layer has a thickness of 30 m andwas introduced to represent the Quaternary clastic complexand the eluvium/colluvium overlying the bedrock. In real-ity, the thickness is not homogeneous and has an inverserelationship with the slope angle; indeed, eluvium/colluvium is usually gravitationally mobilized on slopessteeper than 30°, leaving the bedrock exposed. A betterdefinition of the eluvium/colluvium thickness would im-prove the model, defining the small perched aquifers thatfeed the ephemeral mountain streams. The role of theseaquifers has been omitted from this regional-scale study.

Fig. 6 Simulated hydraulic head distribution of Shenzhen, the resultfrom the top layer. Key to the legend: (1) groundwater flow direction,(2) main river considered in the simulation (RIV package), (3)

Applied General Head Boundary (GHB package); (4) single/clustered measured water levels

Table 2 Estimated groundwater balance from the Shenzhen numericalmodel

Parameter Value

Area [km2] 2,015.8

Total rainfall [m3/year] 4,031.6 × 106

Average groundwater discharge [m3/year] 485.6 × 106

Groundwater from urban recharge [m3/year] 211.6 × 106

192 Hydrogeol J (2019) 27:183–194

Conclusions

A hydrogeological conceptual model of the Shenzhen area(Figs. 2 and 3) and numerical analysis identified and quan-tified the Shenzhen groundwater resources (Fig. 6).Despite a lack of measurements of hydraulic conductivityand recharge, the model fits well to the observed head data.The Quaternary coastal plain, alluvial valleys and eluvium/colluvium blankets represent the main aquifer. Tropicalhumid conditions have created the eluvium-colluviumblanket, which facilitates infiltration. The dense vegetativecover, in the nonurban areas, increases the porosity of thesoil which slows the runoff and also facilitates infiltration.Bedrock is a regional aquitard with relatively unimportantcirculation. According to the conceptual model watershedsand groundwater basins coincide. The numerical simula-tion based on this conceptual model constrained the hy-draulic conductivity and recharge dataset.

Over 40% of the land area of Shenzhen is denselyurbanized, and in this area, infiltration from precipitation issmall but leakage from water mains, sewers and irrigationexcess provides a significant source of groundwaterrecharge. Estimated urban recharge from model calibrationis approximately 250 mm/year, a value which is significant,and consistent with Lerner (2002) who found that in urbanareas, recharge is equal to or higher than in natural areas. Innatural areas, recharge varies as a function of the outcroppinghydrogeological units.

The conceptual model and associated numerical simula-tion provide a comprehensive framework for understand-ing groundwater conditions in Shenzhen and supportingfuture management of the city. The shallow aquifer is im-portant in supporting streams which are environmentallysignificant. A sustainable development and managementof the water resources is essential for the Shenzhen areaand this study represents an important first step for pollu-tion remediation of the aquifers, and the water-table resto-ration. An adaptation of the model could be useful to ana-lyze the Bsponge city^ concept. The same approach of ablock-structured hydrogeological model and calibrateddataset can be reapplied to describe groundwater flow inother areas of Guangdong and nearby Fujian provincewhich have similar geological and climate features, andhigh population density.

Acknowledgements We would like to thank the editors Dr. Jean-MichelLemieux and Dr. Rebecca Doble for the detailed and helpful reviewsprovided. The constructive comments of Dr. Tao Cui and an anonymousreviewer are acknowledged.

Funding information This work has been supported by the National KeyR&D program of China, project no. 2016YFC0402806. Additional sup-port was provided by the National Natural Science Foundation of China(grant no. 41330632).

Open Access This article is distributed under the terms of the CreativeCommons At t r ibut ion 4 .0 In te rna t ional License (h t tp : / /creativecommons.org/licenses/by/4.0/), which permits unrestricted use,distribution, and reproduction in any medium, provided you give appro-priate credit to the original author(s) and the source, provide a link to theCreative Commons license, and indicate if changes were made.

References

Allen MB, McDonald MIB, Zhao X, Vincent SJ, Brouet-Menzies C(1997) Early Cenozoic two phase extension and Late Cenozoic ther-mal subsidence and inversion of the Bohai Basin, northern China.Mar Petrol Geol 14:951–972

Brooks RH, Corey AT (1966) Properties of porous media affecting fluidflow. J Irrig Drain Div Am Soc Civ Eng 92:61–87

Caine SC, Evans JP, Foster CB (1996) Fault zone architecture and per-meability structure. Geology 24:1025–1028

Campbell SDG, Sewell RJ (1997) Structural control and tectonic settingof Mesozoic volcanism in Hong Kong. J Geol Soc 154:1039–1052

Campbell SDG, Sewell RJ (1998) A proposed revision of the tectonicstratigraphy and related plutonic classification of Hong Kong. JGeol Soc Hong Kong 4:1–11

Chang NB, Lu JW, Chui TFM, Hartschorn (2018) Global policy analysisof low impact development for stormwater management in urbanareas. Land Use Policy 70:368–383

Chen J, Chang K, Karacsonyi D, Zhang X (2014) Comparing urban landexpansion and its driving factors in Shenzhen and Dongguang. HabInt 43:61–71

Cilona A, Aydin A, Johnson NM (2015) Permeability of a fault zonecrosscutting a sequence of sandstones and shales and its influenceon hydraulic head distribution in the Chatsworth formation,California, USA. Hydrogeol J 23:405–419. https://doi.org/10.1007/s10040-014-1206-1

Di Salvo C, Moscatelli M, Mazza R, Capelli G, Cavinato GP (2014)Evaluating groundwater resources of an urban alluvial area throughthe development of a numerical model. Environ Earth Sci 72:2279–2299

GMS (2017) GMS usermanual tutorial 10.4 version, groundwater model-ing system. Available at http://gmsdocs.aquaveo.com/GMS_User_Manual_v10.3.pdf. Accessed September 2018

Hao P, Sliuzas R, Geertman S (2011) The development and redevelop-ment of urban villages in Shenzhen. Hab Int 35:214–224

Harbaugh AW, Banta ER, Hill MC, Mc Donald MG (2000) MODFLOW2000, the U.S. geological surveys modular groundwater model: userguide to modularization concepts and the groundwater flow process.US Geol Surv Open-File Rep 00-92, 191 pp

Harbaugh AW (2005) MODFLOW 2005 – The U.S. Geological Surveymodular groundwater model: the groundwater flow processes. USGeol Surv Techniques Methods 6-16A, 253 pp

Held I, Wolf L, Eiswirth M, Holtz H (2006) Impacts of sewer leakage onurban groundwater. In: Hellan JH et al. (eds) Urban groundwatermanagement and sustainability. NATO Science Series IV, Springer,Heidelberg, Germany, pp 189–204

Hong W, Guo R (2017) Indicators for qualitative evaluation of the socialservices function of urban greenbelt systems: a case study ofShenzhen, China. Ecol Indic 75:259–267

Hunt RJ, Zheng C (2012) The current state of modeling. Ground Water50:330–333

Kruse E, Carol E, Mancuso M, Laurecena P, Deluchi M, Rojo A (2013)Recharge assessment in an urban area: a case study of La Plata,Argentina. Hydrog J 21:1091–1100

Hydrogeol J (2019) 27:183–194 193

Lerner DN (1990) Groundwater recharge in urban areas. Atmos Environ24B(1):29–33

Lerner DN (2002) Identifying and quantifying urban recharge: a review.Hydrogeol J 10:143–152

Li ZX, Li XH (2007) Formation of the 1300-km-wide intracontinentalorogeny and post-orogenic magmatic province in Mesozoic SouthChina: a flat slab subduction model. Geology 35:179–182

Li T, Li W, Zhenghang Q (2010) Variations in ecosystem service value inresponse to land use changes in Shenzhen. Ecol Econ 69:1427–1435

Li W, Li L, Qiu G (2016) Energy consumption and economic cost oftypical wastewater treatment systems in Shenzhen, China. J CleanProd 163:374–378

Lu Y, Sun J (1991) An investigation on geomorphology and quaternarydepositions for activity of the Shenzhen fault zone, Guangdongprovince. Seismol Geol 13(2):138–146

Niswonger RG, Prudic DE, Regan RS (2006) Documentation of theunsaturated zone flow (UZF1 package) for modeling unsaturat-ed flow between the land surface and the water table withMODFLOW-2005. US Geol Techniques Methods, Book 6-19,62 pp

Niswonger RG, Panday S, Ibaraki M (2011) MODFLOW-NWT, aNewton Formulation for MODFLOW-2005. US Geol TechniquesMethods 6-A73, 44 pp

Pang C, Li ZX, Xu YG, Wen S, Krapez B (2016) Climatic and tec-tonic controls on Late Triassic to Middle Jurassic sedimentationin northeastern Guangdong Province, China. Tectonophysics677–678:68–87

Passerello MC, Sharp J, Pierce A (2012) Estimating urban-induced arti-ficial recharge: a case study for Austin, TX. Env Eng Geosci 18:25–36

Qin H, Tan X, Fu G, Zhang Y, Hung Y (2013) Frequency analysis ofurban runoff quality in an urbanizing catchment of Shenzhen, China.J Hydrol 496:79–88

Ran G, Eyal S, Yoseph Y, Amir S, Noam W (2013) The permeability offault zones: a case study from the Dead Sea rift (Middle East).Hydrogeol J 22:425–440. https://doi.org/10.1007/s10040-013-1055-3

Reefsgaard JC, van der Slujis JP, Brown J, van der Keur P (2006) Aframework for dealing with uncertainty due to model structure error.Adv Water Resour 29:1586–1597. https://doi.org/10.1016/j.advwatres.2005.11.013

Refsgaard JC, van der Sluijs JP, Højberg AL, Vanrolleghem PA (2007)Uncertainty in the environmental modelling process: a frameworkand guidance. Environ Model Softw 22:1543–1556

Shenzhen Committee (2016) The plan factors and censor for Sponge Cityin Shenzhen (in Chinese). https://wenku.baidu.com/view/df529c634a35eefdc8d376eeaeaad1f346931198.html. AccessedSeptember 2018

Sun J, Jian J, Zhan W, Yi S (2007) A study of the tectonic activity of theShenzhen fault. Adv Earth Sc 22(3):234–240

Wang Y, Xia H, Fu J, Sheng J (2004) Water quality changes in reservoirsof Shenzhen, China: detection using LANDSAT/TM data. Sci TotalEnviron 328:195–206

Wang X, Li H, Yang J, Zheng C, Zhang Y, An A, Zhang A (2017)Nutrient inputs through submarine groundwater discharge in an em-bayment in Daya Bay, China. J Hydrol 551:784–792

WRBSM (2016) Shenzhen Water Resources Bulletin (in Chinese).Water Resources Bureau of Shenzhen Municipality, Shenzhen,China, 18 pp

Xia S, Qiu X, Tong VCH, Xu HX (2011) Three-dimensional seismicimaging of the littoral fault zone in the Hong Kong region.Geology 40:59–62

Xu C, Wang H, Luo Z, Ning J, Liu H (2015) Multilayer stress fromgravity and its tectonic implication in urban active fault zone: a casestudy in Shenzhen, South China. J Appl Geophys 114:174–182

Yang Z, Yi S,WangX (2007) Hydrogeological map of Shenzhen, scale 1:50.000. Shenzhen Geological Construction, Shenzhen, China

Yang C, Sun Y, Zhao X (1991) An assessment of the crustal stability ofShenzhen municipality (in Chinese). Geol. Publ. House, Beijing,China

Zhang GW, Guo AL, Wang YJ, Li SZ, Dong YP, Liu SF, He DF, ChengSY, Lu RK, Yao AP (2013) Tectonics of South China continents andits implications. Sci China Earth Sci 56:1804–1828

194 Hydrogeol J (2019) 27:183–194